Development of a Cognitive Training System Based on Decoding of Single-Trial Event-Related Potentials

Abstract

Our final goal is to promote brain health by developing a brain-machine interface (BMI)-based cognitive training device. We have been developing a cognitive level assessment device called the "Neurotrainer," which uses event-related potential (ERP) analysis. This device was operated by decoding single-trial ERPs using pattern recognition. Here, we extended our method to decoding single-trial ERPs during a racing game with 8 choices of characters. The decoding accuracy of the target character was about 54% for a single trial and 83% for a cumulative total of five trials (chance level = 12.5%). Also, the ERP responses were stronger in the feedback condition than in the no-feedback condition. These results suggest that the BMI might be an effective tool for cognitive training, in which real-time neurofeedback affects the activation of the brain of the players.